Associate Professor at Eindhoven University of Technology
Department of Industrial Engineering & Innovation Sciences

To Bot or Not to Bot: Customer Responses to AI-Augmented Service

With the advent of frontline robots as autonomous interfaces that can interact with and deliver service to customers, many questions arise for firms: Should I replace employees by robots? Does a robot help my organization to be customer oriented? How would customers respond to hybrid teams consisting of at least one employee and one robot collaborating in service provision?

In this talk, we will review current academic knowledge on how customers consider advice and service of AI agents, robots, and human employees.

PhD at Eindhoven University of Technology
Department of Industrial Engineering & Innovation Sciences

From Fit Disruption to Fit Construction for Effective Human–AI Collaboration

AI grants employees unprecedented autonomy over when, how, and for which tasks it is used, yet in practice, many struggle to integrate it in ways that support effective human–AI collaboration. Our study reveals how such collaboration can be achieved: by training employees to actively craft fit between their needs and AI, and to optimize job demands.

Associate Professor at Eindhoven University of Technology
Department of Industrial Engineering & Innovation Sciences

Your People Are Using AI. That Should Worry You

AI adoption is up, but so is overconfidence. Research at TU/e shows that the people most convinced they are using AI effectively are often the ones making the most consequential mistakes: Trusting outputs they cannot verify, missing errors they do not know to look for, and producing work they cannot stand behind when it matters.

Some mistakes can end careers. Fabricated sources in a published report. A flawed analysis that ships with a product. A compliance document nobody checked because the AI sounded confident. And unlike most productivity failures, these are invisible until they are catastrophic by which point the time, money, and reputation are already gone. The uncomfortable finding from TU/e's research is that more AI use without better AI governance does not accelerate productivity. It accelerates mistakes at scale, at speed, across your entire organization simultaneously.

This presentation draws on TU/e's work on AI literacy and organizational capability-building to argue that the leaders who will win in the next three years are not the ones who adopted AI fastest. They are the ones who built the systems (policies, tools frameworks, accountability structures) that make AI use trustworthy. For Brainport's entrepreneurs and business leaders, that is both the risk to manage and the opportunity to seize.

Associate Professor at Eindhoven University of Technology
Department of Industrial Design

Keep calm and trust AI

We keep asking whether we can trust AI. Yet a large language model, on its own, can't reliably subtract 7.11 from 7.8, count the r's in "strawberry," or stop itself from confidently promising a local grocery shop a billion euros in profit. So why does it feel so useful? This talk argues for a shift in perspective: a language model is not the intelligence in the room, instead it is an interface, and possibly the best one ever created. Instead of hunting through menus and buttons, you simply type what you want, and the model translates it into calls to real, trustworthy tools that do the actual work.

Through a live demonstration, I show how a language model wired to the right tools turns confident nonsense into honest, dependable answers. The conclusion is practical: you can trust the AI exactly as far as you trust the tools behind it. So the real skill is choosing and trusting your tools well.

Assistant Professor Cognitive Science and Artificial Intelligence at Tilburg University

Supply Chain Engineering lead at ASML and a PhD at Eindhoven University of Technolog

Effective Human-AI Decision Collaboration

How should humans and AI collaborate in decision making? Drawing on three empirical studies from an inventory planning context, this talk investigates why human planners override algorithmic recommendations.

Some overrides reflect rational preference differences, while others reveal systematic biases. Furthermore, we find that automating routine tasks is not always beneficial and show that human-executed routines may outperform automated ones.

Together, these findings offer a behavioral and organizational perspective on human-AI collaboration: why human involvement sometimes adds value, why it sometimes hurts, and how roles and responsibilities should be carefully designed to realize value potential.

Associate Professor at Eindhoven University of Technology
Department of Industrial Design

Assistant Professor at Eindhoven University of Technology
Department of Mathematics and Computer Science

When Accuracy Isn't Enough: Neuro-Symbolic AI for High-Stakes Decisions

An AI-model can be accurate on average and still be unusable. In banking, healthcare, and public services, a prediction has to respect the rules that govern the process, hold up when the data is thin or unrepresentative, and be explainable to someone who has to sign off on it. Standard neural models, including today's large language models, optimize for average-case performance and offer none of these guarantees.

Neuro-symbolic AI takes a different route: it feeds domain knowledge, such as business rules, service-level agreements, regulations, and clinical guidelines, directly into the learning process. This talk shows what that delivers on real enterprise event logs, from loan applications to hospital intensive care. Injecting rules produces significant accuracy gains precisely where historical data is too sparse to teach them, and keeps predictions compliant even when the training data barely reflects the rules.

Associate Professor at Eindhoven University of Technology
Department of Built Environment